Research on technology evolution analysis method from the problem–solution-effect (PSE) three-dimensional perspective

Abstract Technological evolution analysis is critical to identify technology opportunities and develop innovation strategy. Existing approaches represent patents as keywords, topics or syntactic triples, overlooking the causality of the inventive process. This paper introduces a PSE (Problem–Solution-Effect) framework that considers technology evolution as the dynamics of technical problem elements, solution elements and effect elements. To operate the PSE framework, we propose a PSE-based approach for technological evolution analysis, by identifying the PSE elements through Topic Model from related patent sentences extracted by large language models, constructing the PSE heterogeneous network based on co-occurrence relation among PSE elements, and measuring PSE pathways by evolution metrics to identify emerging, hot and mature pathways. An empirical study in the field of “pre-trained large models” identifies 952 PSE pathways across 28,757 patents, 15 technical problems, 17 technical solutions, and 15 technical effects. We observe solutions addressing “vehicle detection and model redundancy” evolved from simple identification to efficient and intelligent direction, achieving continuous reductions in model complexity, improvements in detection accuracy and real-time performance. The most emerging and hottest pathway (e.g., P2: Complex Medical Decoding Weakness-S3: Multi-Scale Semantic-Prior Fusion-E5: Medical Image Segmentation Optimization), and mature pathways also are identified to help R&D resource distribution. Future research will focus on PSE sub-elements extraction, adaptation to different patent languages and legal jurisdictions, and validation of technological evolution analysis.